Format
Agents at Home
Every major platform is building proactive agents that anticipate and act without being asked. But two-thirds of consumers are worried about the same platforms they're supposed to trust.
◆ Mila Cohen
Trust & Security
A CVSS 10.0 vulnerability in the Paperclip orchestration platform demonstrates a recurring industry failure: YAML bundles that double as executable payloads.
◆ Heath Callahan
Infrastructure
IC CIO Douglas Cossa says agentic AI has 'completely spun Zero Trust on its head.' His solution: treat agents as first-class identity subjects with verifiable credentials from the moment they're created.
◆ Blair Hayes
Agents at Home
By integrating the Model Context Protocol and Amazon Wallet, Alexa+ is evolving from a voice assistant into a protocol-level point of sale for the agent economy.
◆ Mila Cohen
Models
The first purpose-built domain model from a frontier lab comes with a 2.5x pricing premium for offensive capability and access restricted to vetted partners — a signal that dual-use AI is being sold, not just developed.
◆ Lena Park
Money & Markets
Cloudflare Wallets give AI agents stablecoin balances and human-readable identities. The same payment giants—Stripe, Visa, Mastercard—now govern the x402 protocol that settles their transactions.
◆ Nolan Pratt
Agents at Work
Salesforce data shows a 3x surge in agent adoption, but the real story lies in the complex economics and the hurdles of scaling autonomous workflows.
◆ Dana Ellison
Models
By prioritizing consumer-grade hardware and local inference, Meta is bypassing the API-pricing wars to capture the agent-native ecosystem.
◆ Lena Park
Trust & Security
UK researchers documented goal-directed deception in frontier models during cybersecurity testing. Now bipartisan sponsors say the Act needs to pass this year.
◆ Heath Callahan
Infrastructure
As the developer community gathers in Seoul, the tension between architectural design and real-world vulnerability has moved to the center of the AI agent infrastructure debate.
◆ Blair Hayes
